[R-meta] Significance of the individual effects
Cesar Terrer Moreno
ce@@r@terrer @ending from me@com
Mon Sep 24 18:46:05 CEST 2018
Suppose I have a dataset of individual studies with their effect size (yi) and variance (vi) for each individual study. E.g.:
yi <- c(0.296030858, 0.421393627, 0.189653473, 0.014600539, 0.226685055, 0.291373762)
vi <- c(0.0049627734, 0.0130962311, 0.0180149624, 0.0009402604, 0.0134568083, 0.0013970893)
I want to know whether each individual effect is significant or not (i.e. whether the confidence interval overlap with 0).
Thus, I have calculated the lower CI as:
CI.lb <- exp(yi - 1.96*sqrt(vi))
Am I right if I assume: ifelse (yi - CI.lb <= 0, “nonsignificant”, “significant”) ?
More information about the R-sig-meta-analysis